Regression Algorithm of Bone Age Estimation of Knee-joint Based on Principal Component Analysis and Support Vector...

Area of Science:

• Radiology
• Machine Learning
• Biometrics

Background:

• Accurate bone age estimation is crucial for assessing skeletal maturity in adolescents.
• Traditional methods may lack precision, necessitating advanced computational approaches.

Purpose of the Study:

• To develop and validate a machine learning regression model for bone age estimation.
• To apply advanced feature extraction and machine learning techniques to DR knee-joint images.

Main Methods:

• Utilized Histogram of Oriented Gradients (HOG) and Local Binary Patterns (LBP) for feature extraction.
• Applied Principal Component Analysis (PCA) for dimensionality reduction.
• Developed a Support Vector Regression (SVR) model for bone age prediction.
• Employed k-fold cross-validation for model optimization and an independent test set for validation.

Main Results:

• The model demonstrated high accuracy in bone age estimation for both male and female adolescents.
• Achieved accuracy rates of 80.67% (male) and 80.19% (female) within a ±0.8 year error range.
• Reported accuracy rates of 89.33% (male) and 90.45% (female) within a ±1.0 year error range.
• Mean Absolute Error (MAE) was approximately 0.485 years, and Root Mean Square Error (RMSE) was approximately 0.6 years for both genders.

Conclusions:

• The developed model, integrating PCA for feature reduction and SVM for regression, provides a highly accurate method for bone age estimation.
• This approach offers a reliable and precise tool for assessing skeletal maturity in Xinjiang Uygur adolescents using knee-joint DR images.